Prospecting on X (Twitter) from complaints and questions

The case for X: problems announced the moment they happen

X is where people go to complain while they are still annoyed. When a tool breaks, a price rises or a vendor ships something unpopular, the reaction lands on X within minutes. That timing is the whole point. A lead found on X today is often worth more than the same lead found next month, because the frustration that made them post is still there.

It also has an unusual audience for a social network: developers, founders, indie hackers, marketers, journalists, investors and a long tail of niche professional communities (data, design, security, crypto, finance). If your buyer is technical or builds in public, they are probably here.

Who you will not find

Small local businesses, most HR and finance buyers, and large-company managers who keep their professional life on LinkedIn. Plenty of professionals read X and never post. And a big share of accounts are brands, bots, reply-farmers and anonymous personas, which pad every search with noise.

What a buying signal looks like on X

X posts are short, so signals are blunt. Five patterns worth searching:

• Outage and frustration posts: "[competitor] is down again", "third time this week [tool] lost my data".
• Recommendation requests: "anyone recommend a [category] that doesn't cost a fortune", "what are you all using for".
• Switching announcements: "finally moving off [tool]", "cancelled [tool] today".
• Price reactions: replies under a competitor's pricing change ("just doubled our bill").
• Build-in-public pain: founders posting "spent all weekend wiring up [task] by hand".

The replies are often richer than the post. A founder asks "what do you use for X?" and fifty people answer: some recommending tools (vendors and fans), some saying "following, same question" (buyers). Those "same question" replies are leads.

How to do it by hand

1. Build a short list of search queries. X's advanced search supports exact phrases in quotes, OR, excluding words with a minus sign, and filtering to replies or by engagement. A good starting query: "anyone recommend" (crm OR "sales tool") -hiring.
2. Add competitor queries. The competitor's name plus "down", "alternative", "pricing", "cancel".
3. Sort by Latest, not Top. Top shows popular posts, which are usually from accounts with big audiences. You want the ordinary user who posted an hour ago.
4. Check the profile. Bio, pinned post, link in bio. Does the person plausibly buy what you sell?
5. Reply in public with help, not a pitch. If someone's tool is down, a workaround is better outreach than a link to your product.
6. Save the searches and run them every day or two. Old results are mostly worthless here.

Budget fifteen to thirty minutes a day. Missing a week matters more on X than on most sources.

What you can and cannot learn about a person

You can learn: their handle, display name, bio, location field (free text, often fanciful), the link in their bio, and what they post about. Their post history tells you a lot about role and interests.

You cannot reliably learn: their employer (only if stated), an email address, or company size. Many professional accounts state a role ("PM at a fintech") without naming the company. Expect to identify a minority of the people you find.

X's rules on scraping and automation

X's Terms of Service say you "cannot scrape the Services without X's express written permission". Its developer policy, which points to the Automation Rules, tells developers to "always get a user's explicit consent before sending them automated replies or messages", to respect opt-out requests immediately, and to "never perform bulk, aggressive, or spammy actions, including bulk following".

In practice that means:

• Write replies yourself, one at a time, about what the person actually said.
• Do not mass-follow everyone who mentions a competitor in the hope of a follow-back.
• Do not cold DM people who have DMs restricted to followers; it will not arrive, and working around it looks like spam.
• If someone says they are not interested, stop.

Hit rate and the usual failures

We have no honest benchmark to give you; it varies with the market. The common reasons X prospecting produces nothing:

• Searching category words. "CRM" on X returns jokes, job ads and vendor marketing. First-person phrases ("I need a", "we switched") cut most of that.
• Arriving late. A complaint answered three weeks later reads as a vendor trawling search, because it is.
• Pitching under outage posts. Replying "try us instead" under someone's angry post looks opportunistic. A useful fix earns the right to mention a product later.
• Counting anonymous accounts. An account with a cartoon avatar and no bio may be a great buyer, but you will struggle to reach them anywhere else.

How Sluice does it

You add signals (phrases of three words or more) and Sluice searches X for the phrase alone. Adding a city would mostly find people talking about the city, so X searches are global. You can also paste a single post, and Sluice reads the people who engaged with it.

Each person is looked up for role, company and location where a match exists, then scored 0 to 100 against your customer profile, including whether what they wrote suggests they have the problem now. The ones scoring 70 or more are worth contacting, and each links back to the post. Every signal records how many searches it opened, how many people it found and how many cleared 70, so a phrase that only catches noise is easy to spot and switch off. Autopilot (Working plan and up) can run these searches on a schedule, which suits X, where freshness matters.

The honest limits: many X profiles carry no company at all, so those leads rest on their words alone. Sluice drafts a reply from what they said, but it does not post to X for you; you copy it and post it yourself. Access to X data depends on platform terms that can change.

Compare how the same phrase behaves on Reddit, where threads last longer, and on LinkedIn, where profiles are richer. For writing the reply itself, start with buying signals that actually convert.